Global monitoring and forecasting of aerosols are required to analyse and predict the impacts of aerosols on air quality and their role in modulating the climate variability. To achieve this, the Copernicus Atmosphere Monitoring Service (CAMS, http://www.copernicus-atmosphere.eu) provides reanalysis records and operational 5-day forecasts of aerosols using the Integrated Forecasting System (IFS), which combines state-of-the-art meteorological and atmospheric composition models together with the data assimilation of satellite products. The current CAMS aerosol monitoring and forecasting system relies on the assimilation of Aerosol Optical Depth (AOD) at 550 nm NRT observations derived from MODIS (TERRA and AQUA satellites) and PMAp (synergy between GOME-2, IASI, AVHRR onboard METOP-A,B,C satellites) datasets. Implementing new observational data streams is of great importance in order to benefit from enhanced accuracy of new observations, to increase the spatial and temporal coverage of the observations, and to increase the resilience of the data assimilation system to the failure of instruments.
In this paper, we present results from the implementation of two additional AOD products in CAMS, namely the SENTINEL3-A and -B/SLSTR AOD NRT product (collection 2.0) and the VIIRS EPS AOD NRT product from S-NPP and NOAA-20 satellites. The consistency between MODIS, PMAp, VIIRS and SLSTR AOD products as well as their differences with the modelled AOD were evaluated over a 6-month experiment, from December 2019 to June 2020, at both global and regional scale. The evaluation was done at the model grid scale to understand how the differences between products may impact the assimilation. Several assimilation experiments were designed to test distinct assimilation strategies in terms of bias correction and choice of an anchor.
The comparison of satellite AOD shows an overall good consistency between VIIRS and MODIS. VIIRS AOD is frequently lower than MODIS over the ocean background and higher over biomass burning and desert land regions. All the products capture the Mid-Atlantic dust outbreak and the smoke plume from Central Africa but VIIRS resolves finer spatial structures such as the transport of Australian biomass burning aerosols over the Pacific that are not detected by MODIS and the model. S-NPP has systematic higher AOD than NOAA20 AOD which is related to known calibration uncertainties over ocean. SLSTR AOD shows much smaller values than the rest of the products which is mainly related to differences in representativity due to the stringent cloud filtering applied to the SLSTR product. The largest diversity between the products is observed over the Southern Ocean where surface roughness and white foams induced by strong winds are important sources of uncertainties for aerosol retrievals.
The assimilation of VIIRS and SLSTR AOD slightly reduces the bias in the AOD forecast evaluated against AERONET. It leads to a reduction of the analysis increment over ocean which was supposed to be too high due a positive bias in the MODIS/TERRA AOD product. The assimilation of VIIRS leads to an increase of the increments over biomass burning and desert regions. Besides, the assimilation of VIIRS results in an increase of PM2.5 forecast and a reduction of the bias evaluated over Europe.
From the outcomes of this work, strategies for assimilating multi-satellite AOD products in atmospheric composition global models are discussed and recommendations are given for the development of future satellite AOD products.

Within ESA’s Climate Change Initiative (CCI), the Aerosol_cci+ project (3/2018 – 3/2022) focuses on evolving, qualifying and applying aerosol retrieval algorithms for the dual-view radiometer sensor line covering the periods 1995 – 2012 and 2016 – 2030 (ATSR-2 onboard ERS-2, AATSR onboard ENVISAT; SLSTR onboard SENTINEL-3A and SENTINEL-3B). Aerosol_cci+ builds on the legacy of a decade of Aerosol_cci(2) projects since 2010 and works to benchmark algorithms for future operational use in the Copernicus Climate Change Service. Within Aerosol_cci+ limited test datasets are processed to allow a systematic evaluation (by DLR and FMI) comprising of validation against ground-based AERONET measurements and inter-comparisons to other satellite aerosol datasets.

The project works with two algorithms of different maturity to provide Aerosol Optical Depth (AOD) and Fine Mode AOD: A mature algorithm from Aerosol_cci2 (by Swansea university) has been improved to overcome some of its core weaknesses (e.g. over bright surfaces). A second innovative algorithm CISAR (by Rayference) has been adapted to SLSTR and allows combined retrievals of aerosol and clouds without relying on an external cloud mask. Work is continuing to further reduce the over-estimation of total AOD and Fine Mode fraction by the Swansea algorithm and to improve the robustness of the probabilistic aerosol-cloud discrimination in CISAR.

Two specific user case studies have been conducted in Aerosol_cci+ to assess the potential and limitations of the improved datasets. Aerosol_cci+ furthermore supported community inter-comparison experiments and model evaluation (Schuttgens, et al., 2020 and 2021; Gliß et al., 2021; AEROSAT merged dataset, Sogacheva, et al., 2020).

The first study (by ECMWF) tests the impact by data assimilation of the SU SLSTR/3A dataset into the CAMS global atmospheric model for climate services and re-analysis. This study shows the high skill level of the model (even without satellite data assimilation) and assesses the impact of SLSTR assimilation on AOD and PM modelling. Analysis so far shows that further bias reduction for the SLSTR retrieval is needed.

The second study (by MPI) evaluates the use for climate science applications through calculating radiative forcing aerosol direct and indirect effects with an off-line radiative transfer model. Using one test data year from each sensor in the dual-view family (ATSR-2: 1998, AATSR: 2008, SLSTR/3A and SLSTR/3B: 2019 and 2020), the radiative forcing impact is analysed in decadal steps. Also, this study shows the need to reduce the SLTR AOD and Fine Mode AOD biases. This study also explores the ground for a new satellite-based climate indicator “radiative cooling offset by aerosol (and clouds)” in future support of the implementation of the Paris agreement.
The paper will summarize algorithm improvements and their validation and then focus on the two user case studies and further applications of Aerosol_cci datasets.

In this study, we aim to use aerosol observations from both low earth orbiting (LEO) and geostationary (GEO) satellites, measurements from ground-based networks and aircrafts, and atmospheric chemistry and transport models to address the following questions:

1) What are key factors modulating the relationship between column aerosol optical depth (AOD) and surface PM2.5 over different spatial and temporal scales, including aerosol vertical distributions, composition/size, and meteorological conditions?
2) What are the most scientifically robust and logistically feasible ways to convert satellite observations of aerosols to surface PM2.5 for air quality applications?

We will use the global AOD data from Terra, Aqua, and S-NPP satellites and the regional aerosol data from GOES-16 & 17 over North America and Himawari-8 over Asia, which will be complimented with the ground-based and aircraft data for aerosol speciation, size, and composition. The data will be analyzed with the NASA global model GEOS and regional model NU-WRF (NASA Unified WRF) for understanding the physical chemical processes that determine the different AOD-PM2.5 at different spatial and temporal scales and various environment. At the end, we will recommend the practical use of the GEO-LEO data for air quality research and applications.

Dust is the most abundant continental aerosol by mass in the global atmosphere and plays a key role in the Earth system. An accurate quantification of its spatial and temporal distribution is crucial to correctly characterize the effect that dust has on the earth’s energy balance as well as to improve the skill in dust forecasting. High accuracy in the characterization of dust can be best achieved by optimally combining observations and model simulations, through data assimilation. Data assimilation has in fact impacted significantly the forecast and monitoring of aerosols, and is operational in some of the main aerosol forecasting centers. The majority of the aerosol forecasts and reanalyses rely on the assimilation of total aerosol optical depth (AOD), i.e. including all aerosol species. Assimilating AOD may not necessarily constrain well individual aerosol components (and hence dust) since the distribution of the analysis increments among the different species is typically determined by the model first-guess.

This talk provides an overview of recent activities at the Barcelona Supercomputing Center involving the assimilation of satellite dust aerosol products for forecasting and monitoring applications. We report on data assimilation simulations specifically targeting dust aerosols which are performed with the Multiscale Online Nonhydrostatic AtmospheRe CHemistry (MONARCH) model coupled to an ensemble-based technique known as local ensemble transform Kalman filter (LETKF). Ensemble-based techniques use flow-dependent model error amplitudes and structures which evolve during forecast. They are therefore able to capture better instabilities in the background flow compared to techniques requiring pre-calculated, and constant in time, model error structures. We describe the different options of designing our ensemble simulations considering model uncertainties with respect to surface winds, soil humidity, and vertical flux distribution at emission sources and meteorological drivers.

We showcase the use of dedicated dust observations, both total column optical depth (e.g., from MODIS, VIIRS and IASI) and extinction coefficient profiles (e.g., CALIOP-based LIVAS retrievals), to improve dust forecasting and to produce a dust reanalysis over a multi-year period.